After 2 articles were selected for SigGraph in July this year, Style3D has 2 more articles selected for SigGraph Asia!
Joining Google, Adobe, NVIDIA, Intel, Disney, Tencent, and other companies**, it can be said that Style3D's research in the field of computer graphics has ranked among the highest levels of global enterprises.
Recently, at the 16th International Conference on Computer Graphics and Interactive Technologies in Sydney, Siggraph Asia 2023, colleagues from Style3D presented two physics-based fabric simulation technologies**. SIGGRAPH represents the highest level of international heavyweight CG event in computer graphics, and SIGGRAPH Asia is the only chapter of SIGGRAPH. This time, Style3D's two ** articles were selected at the same time, which shows its heavy gold content.
The two first-class research results selected this time will significantly improve the real effect and efficiency of fabrics.
The first article is about improving the stability of dihedral bending models by using an analytical feature system and adaptive geometric stiffness.
Style3D proposes a concise expression of the Hessian matrix of the dihedral angle bending model and an expression of the analytic feature system, revealing that the dihedral angle bending model usually has four positive, negative, and zero eigenvalues. With these expressions, indefinite Hessian matrices can be parsely modified to positive semidefinite matrices efficiently. A variable geometric stiffness matrix for correcting dihedral bending models by orthogonal geometric stiffness matrices is also proposed.
In addition, it is recommended to adjust the compressive stiffness according to the Kirchhoff-Love sheet theory to avoid overcompression. Style3D has developed a method that not only ensures positive semi-qualitativeness, but also avoids instability caused by large bending forces due to degraded geometries. To demonstrate the advantages of the method, in a challenging example, the effectiveness of the existing method in cloth and sheet simulation is compared in experiments.
In flexibility, the curvature of the fabric will determine the realism of the fabric, different physical fabrics have different physical properties, and the "folding clothes" that are easy to achieve in reality are countless calculation results in the digital world. How to stably restore the bending properties, and even in extreme situations (such as large bending forces), it can still perfectly present the real bending feeling, which is a difficult problem for graphics to overcome. Style3D's technological innovation will be applied to Style3D's simulation engine in the near future to enhance the curvature of clothing, providing users with a more realistic and better experience.
Part 2 describes a way to improve the efficiency of cloth simulation on GPUs.
Style3D proposes an efficient method for cloth simulation in the Projective Dynamics (PD) framework. This method combines the advantages of subspace integration and parallel iterative relaxation methods. At its core, it is the use of subspace to accelerate the convergence efficiency of Jacobi-PD, and the clever use of precomputational decomposition to solve the global system in the subspace. The pre-computed decomposed subspace system matrix is utilized in the subspace lbfgs. The LBFGS method uses the system matrix of the subspace as the initial Hessian approximation, combined with Jacobi iterations in the whole space to capture high-frequency details.
This method can be performed efficiently on GPUs, and experiments have shown a significant performance improvement in high-resolution cloth simulations. Subsequently, Style3D will successively implement this method in Style3D's self-developed simulation engine to enhance the overall user experience by providing faster simulation speed. The application of this technology will also once again improve the advantages of Style3D in the speed of cloth simulation and build a higher technical barrier related to Style3D.
See the truth in the subtleties, listen to the thunder in the silence", the improvement of the subtle realism of the fabric is behind the results of countless technicians who have studied day and night. Flexibility**As one of the most difficult and complex fields in graphics, Style3D has been struggling with this field for many years, and this year it has published a total of 4 articles** in Siggraph and SigGraph Asia, which has become an important step in the advancement of related technologies around the world.
As the owner of technical achievements, Style3D will also apply the research results to products such as Style3D Studio, Atelier and Simulator as soon as possible, so that the technology can bring practical application value to the digital research and development of clothing. Let's look forward to more scientific research breakthroughs in the future!
Style3D Studio: 3D clothing design software applied in the textile and garment industry, with its own flexible technology, making 3D clothing design more real, faster and more efficient.
Style3D Atelier: A professional-grade digital apparel modeling tool for the CG industry, making it easy to create, simulate, and present high-fidelity digital garments.
Style3D Simulator: A hyper-realistic clothing plugin for CG industry applications, which can be used for any UE project and improve the efficiency of clothing animation production.
About Style3D-Lingdi Research Institute.
Style3D's Lingdi Research Institute was led by Dr. Wang Huamin, the world's top grapher and former tenured professor at Ohio State University, and the core team members are from Stanford University, University of Utah, Tsinghua University, Zhejiang University and other internationally renowned universities. Lingdi Research Institute is composed of talents with a master's degree or above, of which more than 40% are doctors, and is committed to the research of cutting-edge technologies in the field of flexible body and AI. Lingdi Research Institute has published high-level articles in the world's top journal of computer graphics, Siggraph, and Lingdi Technology has taken the lead in research fields such as computer graphics and AI apparel industry applications. Recently, it has attracted internationally renowned computer graphics experts Yang Yin and Jiang Chen Fanfu to join, and has become the world's leading gathering place for graphics talents.
4 articles were selected**:
fast gpu-based two-way continuous collision handling (siggraph)
second-order stencil descent for interior-point hyperelasticity(siggraph),stable discrete bending by analytic eigensystem and adaptive orthotropic geometric stiffness(siggraph asia)
subspace-preconditioned gpu projective dynamics with contact for cloth simulation(siggraph asia)